Attendees: Eelco, Ben, Guillaume, Yannik, Théophane, Tom + John later

Round on the current blockers

  • Flake owner check (Théophane) -> Needs to be merged
  • ! syntax for the installables (John)
    • Should we keep it in the “UX” board
    • Is this an extension to the ^installable syntax?
    • (Addendum after John joined)
      • This only extends the “drv path” installable (not the other kinds of installables)
      • John: We'll probably want the UI to be consistent in the long term, but not sure how to do that
    • -> Moved back to ”in progress”
  • Nix repl flakes
    • Mostly ready to go, just a silly test failure to fix first
  • Lazy trees (Eelco)
    • Essentially working, but changes the behavior of subflakes
    • Raising some backwards-compat issues
      • Probably fine, but needs to be spelled out explicitly somewhere
  • Improved error reporting (Guillaume)
    • Working, but needs to be partially rewritten because of conflicts with the now merged --debugger PR
  • Better error message for missing files
    • Will be solved as part of the lazy tree PR
  • Installer script update
    • Still nobody having ownership of the installer (Tom somewhat offered that, but as a low prio thing, so might not happen soon)
    • Edit: (Valentin) I had taken on responsibility some weeks ago, never made progress, and it was quickly overridden in priority by documentation work. I did not bail out explicitly. Sorry. :(
  • Common flake outputs doc
    • Already in the template

Follow-up on the recursive flake inputs overriding

https://github.com/NixOS/nix/issues/6549, implem at https://github.com/NixOS/nix/pull/6550

How to globally override a specific input?

Proposed solution: have a .register boolean attribute on flake inputs, that when true will make the given input be implicitly follow-ed by every dependent flake (on certain conditions)

…

  • John: We could have constraints on the inputs (version bounds? Stuff?)

    • Théophane: That could just be implemented as an assert in the outputs function
    • Théophane: It would be great to have a way to defer the lockfile creation to an external tool

  • John: The issue might be worth an RFC to discuss the details

Proposal: Hide it behind an XP feature flag and make an RFC to settle the details

Agenda for the day

Discarded because Rok was supposed to present (on the progress bar) but couldn't make it. For next time then :)

Select a repo